TWI641828B - Method of characterizing structures of interest on semiconductor wafer and semiconductor metrology system - Google Patents

Abstract

The invention discloses a device and method for characterizing a plurality of structures of interest on a semiconductor wafer. A plurality of spectral signals are measured from one or more sensors of a metrology system according to a plurality of azimuth angles from a specific structure of interest. A spectral difference is determined based on the spectral signals obtained for the azimuths. Based on the analysis of the spectral differences, a quality indicator of the particular structure of interest is determined and reported.

Description

Method for characterizing structure of interest on semiconductor wafer and semiconductor metrology system [Cross Reference of Related Applications]

This application claims a previous application by Thaddeus Gerard Dziura et al. Filed on August 6, 2013, U.S. Provisional Application No. 61 / 862,801 and U.S. Provisional Application No. 61 by Thaddeus Dziura, et. / 943,098, the entire text of these cases is incorporated herein by reference for all purposes.

The present invention relates generally to methods and systems for the characteristics of semiconductor wafers, and more specifically to the characteristics of the quality of printed patterns on a semiconductor wafer.

Light lithography or optical lithography systems used in integrated circuit manufacturing have been popular for some time. These systems have proven extremely effective in precisely manufacturing and forming very small details in products. In some light lithography systems, a circuit image is written on a substrate by transferring a pattern through a light beam or a radiation beam (such as UV or ultraviolet light). For example, a lithography system may include a light source or radiation source that projects a circuit image through a photomask and onto a silicon wafer coated with a material (eg, photoresist) that is sensitive to radiation. The exposed photoresist typically forms a pattern that, after development, masks the layers of the wafer during subsequent processing steps such as, for example, deposition and / or etching.

In a metrology technique, it is possible to collect A critical dimension scanning electron microscope (CD-SEM) image and the pattern quality of each image are detected to determine the characteristics of the quality of the printed pattern (such as a periodic grating) on the wafer. This technique is time consuming (for example, several hours), and the judgment of the quality of the grating can usually be slightly subjective. CD-SEM measurements also failed to provide information on subsurface defect structures.

In view of the above, improved devices and techniques for the characteristics of printed patterns are desirable.

The following presents a simplified summary of the invention to provide a basic understanding of some embodiments of the invention. This summary is not an extensive overview of the invention and does not identify key / critical elements of the invention or delineate the scope of the invention. Its sole purpose is to present some concepts disclosed in the text in a simplified form as a prelude to the more detailed description that is presented later.

In one embodiment, a method for characterizing a plurality of structures of interest on a semiconductor wafer is disclosed. A plurality of spectral signals are measured from one or more sensors of a metrology system based on a plurality of azimuth angles from a particular structure of interest. A spectral difference is determined based on the spectral signals obtained for the azimuth. Determine and report a quality indicator of a particular structure of interest based on analyzing the spectral differences.

In a specific embodiment, determining a quality indication of a specific structure is performed without using a model or extracting quantitative features from the specific structure of interest. In another aspect, the spectral difference is an average difference between a plurality of differences between a plurality of wavelengths and a plurality of azimuth-spectral spectral signals. In a further aspect, the spectral difference value is the highest one of the plurality of difference values between the spectral signals of a plurality of azimuth angles specified by one of the plurality of wavelength ranges. In another aspect, the specific structure is a grating structure. In this embodiment, the theoretical or measured spectral difference is determined for the azimuth of a defect-free grating structure, and the average difference is normalized according to the theoretical spectral difference to determine a number of defects.

In another embodiment, measuring the spectrum comprises: Radiometer to generate a differential model. In a further aspect, a differential model is determined between the best fit between an image and a radially symmetric image with a residual error. Then, based on the differential model, it is determined whether the spectral difference value indicates a thin film or a defective structure.

In another aspect, a self-directed self-assembly (DSA) structure, an underlying non-DSA structure, and a patterned photoresist structure are selected as targets. In another embodiment, a CD-SEM tool is used to collect reference material for quantifying pattern defects. Spectral signals are obtained from the azimuth of the pattern structure of a training set with one of the known pattern defects. A first relationship function between a spectral signal measured according to different azimuth angles and a residual error is determined, and the first relational expression is based on the spectral signal obtained from the azimuth angle based on the self-training set. A second relation function between a residual error and a quantization of a pattern defect is determined based on the reference data. Spectral signals measured from a particular structure of interest according to the azimuth angle are input to a second relationship function to determine a quantification of one of the pattern defects for this particular structure. In a further aspect, the first relational function and the second relational function are based on a data reduction technique applied to one of the spectral signals and residual errors for the training set and the specific structure.

In another method embodiment, at one or more sensors of a metrology system, measurement is performed from a plurality of adjacent positions of a thin film or one of the structures designed to be uniform across the measurement points. A plurality of spectral signals. Determine whether one of the spectral signals is an average signal or a mean signal. One of the standard deviations of each of the spectral signals at each position is determined from the average signal or the median signal. Without using a model or extracting a certain amount of features from a film or structure, a quality indicator for that location is determined and reported based on analysis of the standard deviation for each location of the film and structure.

In an alternative embodiment, the invention relates to a system for detecting or measuring a sample. The system includes an illuminator for generating illumination and illumination optics for directing the illumination toward a specific structure according to a plurality of azimuth angles. The system also includes a collection optic for directing a plurality of spectral signals from a specific structure to a sensor in response to illumination in an azimuth. The system further includes any task configured to perform the operations described above. Which one of the processor and memory. In a specific embodiment, the system is in the form of an elliptical polarimeter, and includes a polarization state generator for generating a polarization state in lighting and a polarization state analysis for analyzing a polarization state of an optical signal. Device. In other embodiments, the system is in the form of a spectroscopic ellipsometer, a Muller matrix spectroscopic ellipsometer, a spectroscope, a spectroscatterometer, a beam profile reflectometer, or a beam profile ellipsometer.

These and other aspects of the invention are further described below with reference to the drawings.

2‧‧‧ Spectroscopic Ellipsometer (SE)

4‧‧‧ samples

10‧‧‧ Beam Profile

12‧‧‧ Beam Profile Reflectometer

14‧‧‧Broadband Reflection Spectrometer

16‧‧‧Deep Ultraviolet Reflectance Spectrometer

18‧‧‧Broadband Spectroscopic Ellipsometer (SE)

28‧‧‧ Optics

29‧‧‧optical element

30‧‧‧ Optics

32‧‧‧optical element

33‧‧‧ Optics

34‧‧‧ Optics

36‧‧‧ Optics

38‧‧‧ Optics

40‧‧‧ Optics

42‧‧‧ Optics

44‧‧‧ Optics

46‧‧‧ Optics

48‧‧‧ processor system

50‧‧‧ Optical Elements

52‧‧‧Optics

54‧‧‧ Optics

56‧‧‧ Optics

60‧‧‧Optics

62‧‧‧optical element

64‧‧‧ Optics

66‧‧‧Optical Elements

70‧‧‧optical element

72‧‧‧ Optics

74‧‧‧optical element

76‧‧‧ Optics

78‧‧‧optical element

80‧‧‧ Optical Elements

82‧‧‧optical element

84‧‧‧ Optics

86‧‧‧Optical Element

92‧‧‧Optical Element

94‧‧‧optical element

96‧‧‧ Optics

98‧‧‧Optics / Rotation Compensator

100‧‧‧ Optics

102‧‧‧optical element

104‧‧‧optical element

104a‧‧‧block copolymer material

104b‧‧‧ block copolymer material

104c‧‧‧block copolymer material

104d‧‧‧block copolymer material

122‧‧‧Guide pattern

202‧‧‧Image

204‧‧‧Image

206‧‧‧DSA line spacing pattern / defective grating / defective DSA pattern

208a‧‧‧Underlying chemical pattern / Underlying guide substrate pattern

208b‧ Underlying chemical pattern / Underlying guide substrate pattern

210a‧‧‧Incomplete line

210b‧‧‧Incomplete line

210c‧‧‧Incomplete line

210d‧‧‧Incomplete line

210e‧‧‧Incomplete line

210f‧‧‧Incomplete line

210g‧‧‧Incomplete line

210h‧‧‧Incomplete line

210i‧‧‧Incomplete line

210j‧‧‧Incomplete line

210k‧‧‧Incomplete line

210l‧‧‧Incomplete line

212a‧‧‧Defective bridging part

212b‧‧‧line

212c‧‧‧line

212d‧‧‧line

212e‧‧‧line

212f‧‧‧line

212g‧‧‧Defective bridging part

212h‧‧‧line

212i‧‧‧line

220‧‧‧Defectless DSA Line Pitch Pattern / Defectless DSA Pattern

222a‧‧‧Second Copolymer Component Line

222b‧‧‧Second Copolymer Component Line

224a‧‧‧Second Copolymer Component Line

224b‧‧‧Second Copolymer Component Line

224c‧‧‧Second Copolymer Component Line

224d‧‧‧Second Copolymer Component Line

226a‧‧‧First Copolymer Component Line

226b‧‧‧First Copolymer Component Line

228a‧‧‧Best width substrate part

228b‧‧‧Best width substrate part

302a‧‧‧DSA pattern

302b‧‧‧DSA pattern

302c‧‧‧DSA pattern

304a‧‧‧column structure

304b‧‧‧Column Structure

304c‧‧‧Column Structure

304d‧‧‧Column Structure

306‧‧‧Top surface CD-SEM image

306a‧‧‧Column image part

306b‧‧‧Column image part

306c‧‧‧Column image part

308‧‧‧Bridged columnar structure

310‧‧‧Bridge lower

314a‧‧‧ disconnected higher columnar part

314b‧‧‧ disconnected higher columnar part

314c‧‧‧ disconnected higher columnar part

602‧‧‧wafer

700‧‧‧ procedure

702‧‧‧ operation

704‧‧‧operation

706‧‧‧operation

708‧‧‧Operation

709‧‧‧operation

710‧‧‧operation

712‧‧‧operation

714‧‧‧operation

800‧‧‧ Method

802‧‧‧ operation

804‧‧‧Operation

806‧‧‧ Operation

808‧‧‧Operation

1100‧‧‧Method

1101‧‧‧Operation

1102‧‧‧Operation

1104‧‧‧Operation

1106‧‧‧Operation

1108‧‧‧Operation

1110‧‧‧ Operation

1112‧‧‧ Operation

1114‧‧‧ Operation

1A to 1D are diagrams of an example DSA procedure for implementing a guide pattern. During a annealing process, a block copolymer material on the guide pattern becomes more and more self-ordered.

FIG. 2 is a bridge defect between a DSA line pitch pattern and a defect-free DSA line pitch pattern.

Figure 3 contains three widely varying DSA patterns and their resulting top-view CD-SEM images.

FIG. 4A shows an ellipsometry spectrum for an alpha signal, where the known grating quality is good.

Figure 4B shows a similar set of spectra for a beta signal, where the known grating quality is good.

Figure 5 shows a similar set of spectra for one of the alphas, where the grating is known to be of poor quality.

FIG. 6 shows the spectral difference (Δ) plotted according to the position on a wafer.

FIG. 7 is a flowchart illustrating a procedure for determining a quality feature of a wafer pattern according to an embodiment of the present invention.

8 is a flowchart illustrating another method for indicating quality on a semiconductor wafer using a plurality of measurements in a measurement site according to an alternative embodiment of the present invention.

FIG. 9 shows a plurality of single measurement sites according to an example embodiment of the present invention. One of the adjacent measurement locations.

FIGS. 10A and 10B show two-dimensional beam profile reflectometry (2DBPR) signal residuals for low-quality wafers and high-quality wafers according to an example embodiment of the present invention.

FIG. 11 illustrates a method for evaluating the quality of a DSA grating by using a differential model according to an alternative embodiment of the present invention.

FIG. 12 illustrates an example weighing and weighing system according to an embodiment of the present invention.

In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. The invention may be practiced without some or all of these specific details. In other instances, well-known program operations have not been described in detail to unnecessarily obscure the present invention. Although the invention will be described in conjunction with specific embodiments, it should be understood that we do not intend to limit the invention to the embodiments.

Preface

Many groups currently use guided self-assembly (DSA) as one of the patterning technologies for advanced nodes. 1A to 1D are diagrams using a DSA procedure of a guide pattern 122. During an annealing process, a block copolymer material on the guide pattern 122 becomes more and more self-ordered. As shown, the block copolymer material 104a is initially disordered relative to the guide pattern 122 in FIG. 1A. After a first annealing duration, a slightly more ordered block copolymer material 104b is formed on the guide pattern 122, as shown in FIG. 1B. After more annealing, a more ordered block copolymer material 104c is formed on the guide pattern 122 in FIG. 1C. Finally, an ordered block copolymer material 104d is formed on the guide pattern 122 during a specific annealing duration in FIG. 1D. Then, one of the copolymer components can be etched (not shown) to obtain a fine grating structure from the remaining copolymer material.

In one of the main types of DSA procedures called "chemical epitaxy", the grating The quality of manufacturing is sensitive to the dimensional parameters of the guide pattern and the chemistry of the material. Defects in these gratings can take the form of randomly ordered structures (especially disordered regions) and sub-surface bridges of one of the DSA materials below the other. FIG. 2 illustrates bridging defects between a DSA line pitch pattern 206 and a defect-free DSA line pitch pattern 220. Specifically, the DSA line pitch pattern 206 is formed on the underlying chemical patterns 208a and 208b having one of the non-optimal widths. Due to the second-best underlying guide substrate patterns 208a and 208b, the DSA pattern is disordered. For example, the defect grating 206 includes a first co-polymer component material patterned into incomplete lines 210a, 210b, 210i, and 210j and 210c, 210d, 210e, 210f, 210g, 210h, 210k, and 210l. The second co-polymer component forms defective bridge portions 212a and 212g and lines 212b, 212c, 212d, 212e, 212f, 212h, and 212i.

In contrast, the DSA pattern 220 is defect-free. Second co-polymer module lines 222a and 222b are formed over the optimal width substrate portions 228a and 228b. Additional second co-polymer component lines (224a to 224d) and first co-polymer component lines (eg, 226a and 226b) are also formed between these substrate portions 228a and 228b. When a CD-SEM is used to image the top surfaces of both the non-defective DSA pattern 220 and the defective DSA pattern 206, the resulting images 202 and 204 are substantially the same. Therefore, the underlying bridge defect of the DSA pattern 206 is not detected.

Figure 3 contains three widely varying DSA patterns 302a, 302b, and 302c. Specifically, the DSA pattern 302a is formed of three columnar structures 304a, 304b, and 304c. The DSA pattern 302b is formed of a bridged columnar structure 308, and the DSA pattern 302c is formed of a bridged lower portion 310 and disconnected higher columnar portions 314a, 314b, and 314c. All three structures 302a to 302c result in the same top surface CD-SEM image 306 having three columnar image portions 306a, 306b, and 306c.

In the case of subsurface defects, the CD-SEM tool ignores the defects without modifying the typical measurement conditions and possible damage to the device. Some people believe that the lack of detection of these sub-substrate defects is a serious obstacle to mass production.

One of the goals of part of the current DSA development process is to determine where a well-formed And poorly formed gratings, and where to form a disordered thin film shape. This difference may depend on the lithographic tool exposure and dose used to pattern the guide layer at the bottom of the DSA film stack. The azimuthal asymmetry in the layer's light response can be used to quantify whether a well-ordered grating is formed.

Example embodiments of characteristic patterns: Certain embodiments of the present invention include devices and methods for spectral data collection and signal processing for characterizing the quality of a structure of interest on a semiconductor wafer without requiring any modeling. Characterizing the quality may include an indication as to whether a structure of interest is of good or bad quality. These techniques allow one of the simpler and faster components to characterize procedural yields across an entire wafer based on the analysis of raw spectral signals. In some embodiments, these techniques (in particular) can be applied to characterize the quality of periodic structures (diffraction gratings) and photoresist or guided self-assembly (DSA) structures. In addition to grating structures, other types of wafer structures of interest can be analyzed for quality. Examples include thin films, periodic and non-periodic structures, and more.

In some embodiments, the spectral measurements are performed at multiple azimuth angles. In another example, multiple spectral signals are obtained from nearby measurement locations.

The spectral signals obtained from the measurement sites can include any pair of the same signal type that can be subtracted from or compared to each other. Example signals include, but are not limited to, any type of scattering measurement, spectroscopic, ellipsometry, and / or reflectometer signals. These signals include: Ψ, △, Rs (complex reflectivity of s-polarized light), Rp (p-polarized light) Complex reflectivity), Rs (| r _s | ² ), Rp (| r _p | ² ), R (non-polarized reflectance), α (spectral "α" signal), β (spectral "β" signal), and Functions of these parameters (such as tan (Ψ), cos (△), ((Rs-Rp) / (Rs + Rp)), Mueller matrix elements (M _ij ), etc.). These signals may alternatively or additionally be changed as a function of incident angle, detected intersection, polarized light, incident azimuth, detected azimuth, angular distribution, phase or wavelength, or a combination of more than one of these parameters. Measure. The signals may also be a feature of a combination of signals, such as an average of any of a plurality of ellipsometer and or reflectometer signal types described above. Other embodiments may use a monochromatic or laser light source in which at least one of the signals can be obtained at a single wavelength (instead of multiple wavelengths). The illumination wavelength can be in any range, starting from the X-ray wavelength and reaching the far-infrared wavelength.

Then, the difference (or standard deviation) between two (or more) spectral signals can be directly analyzed to determine whether a model is not used and no feature parameters such as critical dimension (CD) or Film thickness), measure whether the site has a good or bad quality. That is, the quality indicator of one of the measurement sites can only be based on the difference between two or more measured spectral signals.

In a raster target example, the difference between the spectra can be roughly calculated, and the difference can also be determined at various locations across the wafer. At a position where the spectral difference is higher in a relative sensing, a well-formed grating is more likely to exist. Where the grating structure is poor or even out of order, optical measurement will not detect any strong The azimuth is asymmetric, and the spectral difference between measurements in different azimuth directions will be zero or near zero.

FIG. 4A shows an ellipsometry spectrum of the (α) signal measured at two azimuth angles in the same position, and the known grating quality is good. Similarly, FIG. 4B shows a similar set of spectra for a (β) signal, where the grating is known to be of good quality. The existence of a high-quality grating is demonstrated by the separation between the measured spectra from two azimuth angles. In contrast, Figure 5 shows a set of spectra for (α) where the gratings are known to be of poor quality. The existence of a poor quality grating is demonstrated by the lack of separation between the two measured spectra of the signals at two azimuth angles. Although only the (α) signal is shown for a known poor quality grating, similar results will occur for the (β) signal.

FIG. 6 shows the spectral delta for a wafer with a program change from left to right as the position on a wafer 602 changes. Regions with higher spectral differences between orientations are closely related to the yield map measured for the wafer. For example, some contour areas are shown as having a delta (or difference between signals) greater than 0.060, between 0.050 and 0.60, between 0.050 and 0.040, between 0.040 and 0.030, and so on. this Equal differences can also indicate the quality of the grating.

FIG. 7 is a flowchart illustrating a procedure for determining a quality feature for a wafer pattern according to an embodiment of the present invention. In operation 702, initially, spectral data may be collected from the same measurement location at multiple azimuth angles. In general, two or more azimuths may include any suitable angle, such as angles that are orthogonal to one another (although not required) for increased sensitivity to grating or pattern defects. For example, spectral measurements can be obtained in directions perpendicular to and parallel to the direction of the grating.

The measurement may include any suitable spectral radiation signal, such as a scattering measurement, a reflectometer, or an ellipsometer signal (including the examples described herein). The type of acquired signal can be selected based on the signal sensitivity to the structure of interest. For example, certain wavelengths may be more sensitive to certain structure dimensions. The structure of interest may include any patterned or periodic structure, including 2D and 3D gratings, point arrays, and so on.

After acquiring the spectrum, a spectrum difference may then be determined in operation 704. If there are only two spectra collected for a particular structure of interest, then only the spectral signal can be subtracted to obtain a difference signal. If more than two spectra exist, any number of techniques can be used to obtain a difference signal. If there are more than two azimuths, the spectrum for each pair of azimuths can be subtracted. In one embodiment, the difference is obtained for each pair of orthogonal angles. If the spectrum for each azimuth angle is also obtained according to multiple wavelength ranges, then in operation 706, the highest spectral difference value corresponding to a specific one of the wavelength ranges may be selected as appropriate. In another example, the difference signal (eg, at different wavelengths) may be reduced to a single difference. In the illustrated example, in operation 708, a wavelength average value of the absolute value of the spectral difference value may be determined. That is, the average value of the difference signals for different wavelengths can be determined to obtain an average difference signal.

The quality indication may be defined based on any suitable difference threshold or percentage value, such as compared to an average difference. For example, a difference signal above a predefined threshold may result in a "good" quality indication, otherwise a difference signal below or equal to a predefined threshold may result in a "bad" quality indication. For the example shown in Figure 6, the threshold can be set to A Δ greater than 0.06 is an indication of a good quality grating.

A pattern in the form of a good perfect grating (for example) tends to produce a large difference in the spectral signals measured at different azimuths. As a grating structure becomes more defective, the measured signal difference becomes smaller and resembles a uniform film response, which has a zero difference according to different azimuths. That is, a difference between a maximum difference and zero associated with a perfect grating indicates a defect for a grating or pattern structure.

Then, in operation 709, a quality indication of one of the patterns may be determined and reported based on the spectral difference value. The quality indication may be based on the average wavelength of the spectral differences (from operation 708) or the highest spectral difference for a selected wavelength (from operation 706), as appropriate. The procedure can then be ended or repeated for multiple targets.

Then, the quality determination procedure described above is followed by a calibration procedure. Optionally, a calibration procedure is performed to quantify defects for a structure of interest that is considered to be one of the poor qualities. In the illustrated embodiment, in operation 710, the difference between the theoretical spectrum and a perfect grating may also be determined as appropriate. In operation 712, the measured spectral difference is normalized (or calibrated) according to the theoretical spectral difference. Then, in operation 714, a number of defects may be determined and reported based on the normalized spectral difference. Similar to the quality determination, the number of defects may be optionally based on a wavelength average of multiple spectral measurement differences (from operation 708) or based on the highest spectral difference for a selected wavelength (from operation 706). After performing calibration to quantify the defect, the procedure 700 may also be repeated for multiple spectral differences or the procedure may be ended.

Other measurement structural films for which quality can be determined based on comparison of spectral signals. FIG. 8 is a flowchart illustrating another method 800 for indicating quality on a semiconductor wafer using multiple measurements in a measurement site according to an alternative embodiment of the present invention. FIG. 9 is a diagram of a plurality of adjacent measurement locations of a single measurement site according to an example embodiment of the present invention. Initially, in operation 802, a spectral measurement may be obtained at a plurality of adjacent locations of a same measurement site. Adjacent locations from which measurements were obtained are well adjacent to each other without overlapping the actual measurement area.

An average signal or a median signal for all (or a portion of) the obtained spectral measurements may be determined in operation 804. One of each of the measurements is determined from the average or median standard deviation in operation 806. For each measurement position, in operation 808, a quality indicator may be determined and reported based on the corresponding standard deviation at this measurement position. The above steps can be repeated for multiple locations across the wafer to generate a wafer quality map. A calibration procedure can also be used to extend this procedure to quantify defects, as described further herein.

One measurement site for the technique of FIG. 8 may include any suitable structure or structures of interest, such as a grating or thin film structure, and it is expected that the structures are uniform. For example, one grating that is expected to fill the entire measurement site is uniform at different measurement locations across the region of the measurement site, unless the grating is defective. Likewise, it is expected that a thin film filling the measurement site will have the same thickness (and uniformity) across one of the measurement sites and result in the same spectral signal at different measurement locations in the measurement site. There are more defects in these two examples; there will be larger spectral differences. Therefore, this technique can be applied to normal structures, which can include thin films, 2D and 3D gratings, dot (or any other type) arrays, periodic structures, and so on.

Figures 10A and 10B show two-dimensional beam profile reflectometer (2DBPR) signal residuals (or difference signals) for lower quality wafers (Figure 10A) and higher quality wafers (Figure 10B). Although the examples in FIGS. 10A and 10B use angle-resolved signals, the same method can be applied to wavelength-resolved spectra. Certain techniques of the present invention can be considered "model-free" because assumptions about measured semiconductor patterns are not made. In addition, there is no need to extract quantitative values (such as CDs) of the structure of interest from the measured spectra by comparing model results.

Certain model-less embodiments of the invention can be extended by training techniques for known quality structures, such as known bad and good gratings. An algorithm based on machine learning methods (such as neural networks) can be used to make pre-programmed changes in measured spectral signals and a test structure (such as a DSA pattern) based on a known structure of a training set Related. That is, a predefined or known variation with a training set can be measured to obtain a spectral signal, thereby determining a model for correlating the spectral signal with a characteristic or program variation. After completing machine learning After the operation, features and program-related parameters can be extracted from the spectral signals obtained from the structure of interest with unknown characteristics using this model.

In a further embodiment, various program parameters are modified. To illustrate, when the guide pattern pitch varies within the same wafer or on different wafers, the dimensional properties of the DSA pattern are affected. Alternatively, varying the thickness of the block copolymer layer may be part of a design of experiments (DOE) training set. Program parameter variations (such as annealing temperature) can be used as another way to generate a DOE training set, such as processing different wafers at different annealing temperatures.

After a DOE training set is generated, targets manufactured using different program conditions can be measured to obtain a spectrum (such as a specific signal type from a single measurement site at two azimuth angles or from A specific signal type), and then processing the spectra as part of a training algorithm. To determine dimensional parameters (such as contour characteristics (bottom or top CD, sidewall angles, etc.) for the training set), a reference metrology that can be destructive (e.g., transmission electron microscopy (TEM)) ) Or by atomic force microscopy (AFM) or CD-SEM to characterize these targets from the training set. Known characteristic parameters may only include an indication of poor or good quality, rather than a specific metric value.

Deviations from the ideal or desired DSA pattern can also be characterized, for example, by training, so that the measured signal is related to the intra- or extra-spectral (or good quality compared to poor quality) DSA pattern. For example, using one of the bridged DSA structures can produce asymmetrical signals in corner-resolved or laterally polarized signals that would not be produced by an ideal pattern. The asymmetric level and angle or spectral content of these signals can be related to the density of bridging defects (or other defect types) in the measured area. Depending on the spectra generated by these structures (e.g., these spectra from two azimuth angles or from multiple measurement sites), by using a training method, a feature extractor program can be trained to analyze within the spectrum compared to Out-of-spectrum pattern. Spectra from unknown structures can then be input to a trained feature extractor. Feature extractor output can be used for tuning Program parameters to compensate for process drift that occurs in DSA patterning over time, such as in an advanced process control (APC) system.

In a development and use case, contrary to the production use case, the chemical properties of the block copolymer can also be changed to achieve the best patterned size or roughness properties. Any of the training methods described above can be used to centrally vary and characterize these properties in a DOE (Design of Experiments).

In an illustrative example, the method of FIG. 7 may be implemented in a differential model as described in FIG. 11. For example, a differential model can be determined for a zero-degree azimuth (Az0) about a 90 ° azimuth (Az90): Az0 = f (Az90) + err, where f () makes the Az90 signal best fit to A function of the Az0 signal, and err represents the residual error between the Az90 and Az0 signals. One of the minimum and maximum residual errors for the entire wafer can be determined. When the Az0 and Az90 signals are very similar, then this residual error will be smaller (minErr), at the level of random noise. This smaller minErr corresponds to a poorly formed DSA structure. In contrast, a larger err (maxErr) corresponds to a well-formed DSA structure. The error between minErr and maxErr can be used to evaluate the relative quality of DSA gratings.

FIG. 11 illustrates a method 1100 for evaluating the quality of a DSA grating by using a differential model according to an alternative embodiment of the present invention. Initially, in operation 1101, a CD-SEM or other metrology tool may be used to collect references to DSA defects from a reference DSA structure. That is, CD-SEM can be used to classify a plurality of DSA reference structures such as having a poor or good quality (or defective or non-defective) and a different number of defects. This reference contains quality indications and / or defect counts for known DSA structures and their corresponding spectral data. The spectral reference will contain the same type of signal as collected for the training set below.

In operation 1102, spectral signals at both 0 and 90 ° azimuth angles may be collected from each of a plurality of DSA structures originating from a training set. Although the illustrative examples describe orthogonal azimuths, other angles may be used. However, the azimuth pairs are at different angular positions. Any type of spectral signal (such as spectroscopic data) can be collected. The training set DSA structure will contain different numbers Defects.

Then, in operation 1104, a principal component analysis (PCA) may be performed on the spectral data from the training set at both azimuth angles of 0 and 90 °. Any suitable feature extraction technique (except PCA) can be implemented to extract a feature from a pair of spectral signals with the best information while reducing the data set. Other examples of automated feature extraction techniques include independent component analysis (ICA), local linear embedding (LLE) algorithms, and so on.

A relationship between the PCAs performed for the two azimuths may be determined in operation 1106. In general, this relationship will represent a function that best correlates PCA data from two azimuths. Residual data may then be determined in operation 1108. The residual data is generally the difference between the best-fit functions that correlate the PCA data from the two azimuths. Then, in operation 1110, a PCA may be performed on one or several principal components (such as the first principal component PC1) of the residual data and the PCA residual data that may be selected.

Then, a relationship between the PC1 residual data and the reference data may be determined in operation 1112. For example, the PC1 residual data can be compared with the PC1 data for reference data obtained from a known DSA pattern with known defect qualities (such as poor or good quality and / or many defects), and then, it can be determined that PC1 is related to quality One model. Then, in operation 1114, the DSA pattern quality of the unknown DSA pattern may be determined based on the relationship between the PC1 residual data and the quality determined from the spectrum collected from an unknown DSA.

Although the following example embodiments are described in terms of using a first principal component derived from a PCA transformation to extract information about a quality parameter, other embodiments may utilize other feature extraction results or techniques. For example, the first principal component and the second principal component may be used as determined via the PCA. Any number of principal components can be selected based on the specific needs of the application. In yet another example, the output from another feature extraction tool such as ICA or LLE can be used.

In a PCA embodiment, the extracted features correspond to this new coordinate along which the signal data set is transformed to a different coordinate system and the transformed data set has the most changes. Selection of a specific dimension (or direction or projection direction) of the system, which provides the most information about the quality of the feature. The first principal component corresponds to the transformed direction or dimension of one of the PCA transformed data sets found to have the most variation. The second principal component has a second most variation and so on.

In general, the methods described above can be calibrated to determine quantitative characteristics and / or program parameter values. For example, multiple points measured by CD-SEM can be used to establish the relationship between the number of defects and errors: Number of defects = Cal (err) or Number of defects = Cal (Az0-f (Az90)), where Cal () is a calibration function obtained from a CD-SEM reference measurement site.

The signal used to generate the differential model may be an original signal (such as α, β) or a component of the original signal obtained by PCA, Independent Component Analysis (ICA), Local Linear Embedding (LLE), or other feature extraction methods.

In another illustrative model example, a 2DBPR differential model is generated from the measurement. A 2DBPR image is radially symmetrical (when the sample being measured is similar to a thin film) or non-radially symmetrical (when the sample has some periodic structure). Using this feature, a radially symmetrical surface can be matched with 2DBPR images.

Img = f (Img) + err, where f () is the radial symmetry function of the best-fit image, and "err" is the residual error corresponding to the asymmetric part. An asymmetric portion can mean that the image is at least partially not radially symmetric and is a defective film.

In a similar way as in the case of non-2DBPR, the residual error can be used to evaluate the level of defects, and in the case of calibration, the number of defects can be determined. For non-2DBPR situations, the components of the 2DBPR image can be used instead of the original signal. In both cases, the standard deviation, maximum, median, or error distribution can be used as a measure of DSA grating quality.

In another illustrative model, one technique may include isolating the effects of lower-level changes (de-correlation) when applied to multiple targets. A DSA target can be used as well as one that does not have a DSA layer but has the same underlying layer. The following cases can be used for lower-level decorrelation: 1. Use a combined signal of two or more targets to extract significant components, and 2. Use a differential model for the residual term: residual term = S _dsa -f (S _ul ) Where S _dsa is a signal from a DSA target, and S _ul is a signal from a lower target, and f () is a function that best fits S _ul to S _dsa .

After applying the underlying de-correlation, any of the DSA differential models described above can be applied to DSA defect measurement.

Certain techniques of the present invention can achieve significant time savings in pattern quality for characterizing photoresist or DSA patterns, and reduce the cost of developing and optimizing the patterning process.

Any suitable combination of hardware and / or software can be used to implement any of the techniques described above. In a general example, a weights and measures tool may include a lighting system that illuminates a target, a collection system that captures related information provided by interaction (or lack thereof) with a lighting system of a target, device or feature A system that analyzes and analyzes information collected using one or more algorithms. Metrology tools can generally be used to measure structure and material characteristics (e.g., material composition, dimensional characteristics of structures, and film (such as film thickness) and / or critical dimensions of structures, overlays, etc.) associated with various semiconductor processes. Etc.) of various radiation signals. These measurements can be used to facilitate process control and / or yield efficiency in the fabrication of semiconductor die.

A metrology tool may include one or more hardware configurations that may be used in conjunction with certain embodiments of the invention. Examples of these hardware configurations include (but are not limited to) the following: spectroscopic ellipsometer (SE), SE using multi-angle lighting, SE measurement Muller matrix elements (e.g. using a rotation compensator), single Wavelength Ellipsometer, Beam Profile Ellipsometer (Angular Resolution Ellipsometer), Beam Profile Reflectometer (Angular Resolution Reflectometer), Broadband Reflectance Spectrometer (Spectral Reflectometer), Single Wavelength Reflectometer, Angular Resolution Reflectometer, Imaging Systems and scatterometers (such as speckle analyzers).

The hardware configuration can be divided into discrete operating systems. On the other hand, one or more hard The body configuration is combined into a single tool. An example of combining multiple hardware configurations into a single tool is further illustrated and described in U.S. Patent No. 7,933,026, the entirety of which is incorporated herein by reference for all purposes. FIG. 12 shows, for example, a schematic diagram of an exemplary metrology tool including: a) a broadband SE (e.g. 18); b) using a SE (e.g. 2) of a rotation compensator (e.g. 98); c) A beam profile ellipsometer (e.g. 10); d) a beam profile reflectometer (e.g. 12); e) a broadband reflection spectrometer (e.g. 14); and f) a deep ultraviolet reflectance spectrometer (e.g. 16). In addition, many optical elements (such as 92, 72, 94, 70, 96, 74, 76, 80, 78, 98, 100, 102, 104, 32/33, 42, 84, 60, 62, 64, 66, 30, 82, 29, 28, 44, 50, 52, 54, 56, 46, 34, 36, 38, 40, and 86), including specific lenses, collimators, mirrors, quarters A wave plate, polarizer, detector, camera, aperture and / or light source. The wavelength used in optical systems can vary from about 120 nanometers to 3 microns. The azimuth angle for the optical system can also be changed. For non-elliptical polarimeter systems, the collected signals can be polarized resolved or unpolarized.

Figure 12 provides a diagrammatic illustration of one of a number of metrology heads integrated on the same tool. However, in many cases, multiple metrology tools are used for measurements on a single or multiple metrology targets. Several embodiments of multiple tool metrology are further described in US Patent No. 7,478,019, "Multiple tool and structure analysis" by Zangooie et al., The entirety of which is incorporated herein by reference for all purposes.

Some hardware-based lighting systems can include one or more light sources. One or more light sources can produce light with only one wavelength (e.g., monochromatic light), light with many discrete wavelengths (e.g., polychromatic light), light with multiple wavelengths (e.g., broadband light), and / or continuous or at wavelengths Jump between and sweep light across wavelengths (such as a tunable light source or a swept light source). Examples of suitable light sources are: a white light source, an ultraviolet (UV) laser, an arc lamp or an electrodeless lamp, a laser sustained plasma (LSP) source (e.g., available from Massachusetts) A source commercially available from Enburntiq Technology, Inc. of Woburn), a super source (such as a broadband laser source) (e.g., a source commercially available from NKT Photonics Inc., Morganville, New Jersey), or a shorter wavelength source (Such as an x-ray source, an extreme ultraviolet source, or some combination thereof). The light source may also be configured to provide light with sufficient brightness, in some cases it may be a brightness greater than about 1 W / (nm cm ² Sr). The metrology system can also include fast feedback to one of the light sources to stabilize its power and wavelength. The output of the light source may be delivered via free space propagation or in some cases via any type of fiber optic or light guide.

In turn, one or more detectors or spectrometers are configured to receive illumination that is reflected or otherwise scattered from the surface of the sample 4 via a collection optical element. Suitable sensors include charged coupled devices (CCD), CCD arrays, time delay integration (TDI) sensors, TDI sensor arrays, photomultiplier tubes (PMT), and others Sensor. The measured spectrum or detected signal data (as a function of position, wavelength, polarization, azimuth, etc.) can be transmitted from each detector to the processor system 48 for analysis.

It should be recognized that the various steps described throughout this disclosure may be performed by a single processor system 48 or (alternatively) a multi-processor system 48. Furthermore, different subsystems of the system of FIG. 12 (such as a spectroscopic ellipsometer) may include a computer system adapted to perform at least a portion of the steps described herein. Therefore, the foregoing description should not be construed as limiting the invention, but merely as an illustration. In addition, one or more processor systems 48 may be configured to perform any other steps of any of the method embodiments described herein.

In addition, the processor system 48 may be communicatively coupled to a detector system in any manner known in the art. For example, one or more processor systems 48 may be coupled to a computing system associated with a detector system. In another example, the detector system may be directly controlled by a single computer system coupled to the processor system 48.

The processor system 48 of the metrology system may be configured to receive and / or obtain data or information from a subsystem of the system by a transmission medium that may include one of a wired and / or wireless portion. In this manner, the transmission medium can be used as the processor system 48 and other subsystems of the system of FIG. 12. Data link.

The processor system 48 of the integrated metrology system may be configured to receive and / or obtain data or information from other systems (e.g., measuring spectra, differential signals, statistics) by a transmission medium that may include one of a wired and / or wireless portion. Results, reference or calibration data, training data, models, extracted features or transformation results, transformed data sets, curve fitting, qualitative and quantitative results, etc.). In this manner, the transmission medium can be used as a data link between the processor system 48 and other systems, such as a memory onboard metrology system, external memory, reference measurement sources, or other external systems. For example, the processor system 48 may be configured to receive measurement data from a storage medium (such as internal or external memory) via a data link. For example, the spectral results obtained using the detection system can be stored in a permanent or semi-permanent memory device (such as internal or external memory). In this regard, spectral results can be imported from on-board memory or from an external memory system. Furthermore, the processor system 48 can send data to other systems via a transmission medium. For example, the qualitative and / or quantitative results determined by the processor system 48 may be communicated and stored in an external memory. In this regard, the measurement results can be exported to another system.

The processor system 48 may include, but is not limited to, a personal computer system, a host computer system, a workstation, an imaging computer, a parallel processor, or any other device known in the art. In general, the term "processor system" can be broadly defined to include any device having one or more processors that execute instructions from a memory medium. The program instruction implementation method (such as the methods described herein) may be transmitted via a transmission medium (such as a wire, cable, or wireless transmission link). The program instructions may be stored in a computer-readable medium, such as a memory. Exemplary computer-readable media include read-only memory, a random access memory, a magnetic or optical disk, or a magnetic tape.

Metrology tools can be designed to perform many different types of measurements related to semiconductor manufacturing. Certain embodiments of the invention for determining quality and / or quantitative values may utilize such measurements. Additional metrology techniques for determining specific target characteristics can also be used as described above. Quality judgment technology combination. For example, in some embodiments, the tool can measure the spectrum and determine characteristics of one or more targets, such as quality and number of defects, critical dimensions, overlaps, sidewall angles, film thickness, process-related parameters (e.g., focus And or dose). A target may include certain areas of interest that are periodic in nature, such as, for example, a grating in a memory die. The target may include multiple layers (or films), the thickness of which may be measured by a metrology tool. Targets may include target designs placed (or already present) on a semiconductor wafer for use with, for example, alignment and / or stack-up alignment operations. Some targets can be positioned at various locations on a semiconductor wafer. For example, the target may be positioned within the scribe line (eg, between grains) and / or in the grain itself. In some embodiments, multiple targets are measured by the same or more metrology tools (at the same time or at different times) as described in US Patent No. 7,478,019. Data from these measurements can be combined. Data from metrology tools can be used in semiconductor processes (for example) feedforward, feedforward, and / or lateral feedback correction procedures (such as lithography, etching), and thus a complete process control solution can be produced.

As semiconductor device pattern sizes continue to shrink, smaller weighting targets are often required. In addition, measurement accuracy and matching with actual device characteristics increase the need for device-like targets and measurements in the die and (even) on the device. Various metrology implementations have been proposed to achieve this goal. For example, focused beam ellipsometers based on primary reflecting optics are one of them and are described in Piwonka-Corle et al. (US Patent No. 5,608,526 "Focused beam spectroscopic ellipsometry method and system"). Apodizers can be used to mitigate the effects of light diffraction that cause propagation of illuminating electricity beyond the size defined by geometric optics. The use of an apodizer is described in Norton, "Apodizing filter system useful for reducing spot size in optical measurements and other applications", US Patent No. 5,859,424. The use of high numerical aperture tools that use simultaneous illumination at multiple incident angles is another way to achieve smaller target capabilities. For example, in the US Patent No. "Critical dimension analysis with simultaneous multiple angle of incidence measurements" of Opsal et al. This technique is described in No. 6,429,943.

Other measurement examples may include measuring a composition of one or more layers of a semiconductor stack, measuring certain defects on (or within) a wafer, and measuring the amount of light lithography radiation exposed under the wafer. In some cases, metrology tools and algorithms can be configured to measure non-periodic targets, see, for example, "The Finite Element Method for Full Wave Electromagnetic Simulations in CD Metrology Using Scatterometry" by P. Jiang et al. (U.S. Patent No. 61/830536, KT Invention, page 4063) or "Method of electromagnetic modeling of finite structures and finite illumination for metrology and inspection" by A. Kuznetsov et al. (U.S. Patent No. 61/761146, under application Or KT invention p. 4082).

The measurement of the parameters of interest can also involve many algorithms. For example, the optical interaction with the incident beam of a sample can be modeled using an electromagnetic (EM) solver, and these algorithms can be used as rigid coupled wave analysis (RCWA), finite elements Method (finite element method, FEM), dynamic difference method, surface integration method, volume integration method, finite-difference time-domain (FDTD) and others. A geometric engine or (in some cases) a procedural modeling engine or a combination of both can be used to model the target of interest. The use of program modeling is described in "Method for integrated use of model-based metrology and a process model" by A. Kuznetsov et al. (U.S. Patent No. 61/738760, page 4025). For example, a geometry engine can be implemented in the AcuShape software product from KLA-Tencor, Milpitas, California.

The collected data can be analyzed through many data fitting and optimization techniques and technologies, including: libraries, fast order reduction models; regression; machine learning algorithms (such as neural networks, support-vector machines (support-vector machines (SVM)); size reduction algorithms (such as, for example, principal component analysis (PCA), independent component analysis (ICA), local-linear embedding (local- linear embedding (LLE); sparse representations (such as Fourier or wavelet transforms); Kalman filters; algorithms that facilitate matching from the same or different tool types, and others.

The collected data can also be analyzed by algorithms that do not include modeling, optimization, and / or fitting, such as provisional patent application No. 61/745981, the full text of which is incorporated herein by reference. , And as described in the text.

It is usually best for metrology applications using one or more methods (such as the design and implementation of computing hardware, parallelization, distribution of computing, load balancing, multi-service support, dynamic load optimization, etc.) Optimize the calculation algorithm. Different implementations of the algorithm can be performed in firmware, software, field-programmable gate array (FPGA), programmable optical components, etc.

The data analysis and fitting steps can be used to continue one of the following goals: measurement of quality, number of defects, CD, side wall angle (SWA), shape, stress, composition, film, band gap, electrical Nature, focus / dose, superimposition, generation of program parameters (such as photoresist state, partial pressure, temperature, focus model) and / or any combination thereof; modeling and / or design of metrology systems; and modeling, design, And / or optimization of measurement objectives.

Certain embodiments of the invention presented here generally address the field of semiconductor metrology and program control, and are not limited to hardware, algorithm / software implementations and architectures, and use the cases outlined above.

Although the above invention has been described in some detail for purposes of clarity, it will be understood that certain changes and modifications may be practiced within the scope of the accompanying patent application. It should be noted that there are many alternative ways of implementing the procedures, systems, and devices of the present invention. Accordingly, this embodiment is to be considered as illustrative and not restrictive, and the invention is not limited to the details given herein.

Claims (20)

A method of characterizing a plurality of structures of interest on a semiconductor wafer, the method comprising: measuring one or more sensors from a metrology system, measuring the plurality from a specific position of a specific structure according to a plurality of azimuth angles Spectral signals, where the specific structure is a grating structure; determining a difference signal for the spectral signals obtained for the azimuth angles; if the difference signal is higher than a predefined threshold Value, it is determined and reported that the specific structure does not contain a defect; and if the difference signal is equal to or lower than the predefined threshold or equal to zero, it is determined and reported that the specific structure contains a defect, wherein the defect is low On a surface of that particular structure. The method as claimed in claim 1, wherein, without using a model or extracting a certain amount of feature values from the specific structure, determining whether the specific structure includes or does not include a defect is performed. The method of claim 1, wherein the difference signal is an average difference between a plurality of differences between the spectral signals at the azimuth angles through a plurality of wavelengths. The method of claim 1, wherein the difference signal is the highest of the plurality of differences between the spectral signals of the azimuths specified in one of the plurality of wavelength ranges. If the method of item 3 is requested, the method further comprises: determining a theoretical or measured spectral difference for the azimuths of a defect-free grating structure; and normalizing the average difference by the theoretical spectral difference To determine a number of defects. The method of claim 1, wherein the specific structure is selected from a self-directed self-assembled (DSA) structure, an underlying non-directed self-assembled (non-DSA) structure, or a patterned photoresist structure. A method of characterizing a plurality of structures of interest on a semiconductor wafer, the method comprising: measuring a plurality of spectra from a specific structure of interest from one or more sensors of a metrology system according to a plurality of azimuth angles Signal; determining a spectral difference based on the spectral signals obtained for the azimuths; determining and reporting a quality indicator of the particular structure of interest based on analyzing the spectral difference; using a CD SEM tool to collect quantification References for pattern defects; from the pattern structure of a training set with one of the known pattern defects, to obtain a spectral signal according to these azimuth angles; to determine one of the A relationship function, wherein the first relationship is based on the spectral signals obtained from the training set according to the azimuths; based on the reference data, determining a second between a residual error and a quantization of a pattern defect A relationship function; and inputting the spectral signals measured from the particular structure of interest according to the azimuths to the second relationship function to determine whether the One defect of the pattern structure given quantization. The method of claim 7, wherein the first relation function and the second relation function are based on a data reduction technique applied to the spectral signal and one of the residual errors for the training set and the specific structure. The method of claim 7, wherein measuring the spectrum comprises: generating a differential model by using a two-dimensional beam profile reflectometer. The method of claim 9, further comprising determining a differential model between an image and a best fit of a radially symmetric image with a residual error, and determining whether the spectral difference indicates a thin film based on the differential model. Or defective structure. A semiconductor weighing and weighing system includes: an illuminator for generating illumination; an illumination optics for guiding the illumination toward a specific position of a specific structure according to a plurality of azimuth angles, wherein the specific structure Is a grating structure; the collection optics are used to direct a plurality of spectral signals from a specific structure to the sensor according to the azimuths in response to the illumination; the sensor is used for The azimuths, obtaining the plurality of spectral signals from the specific structure; and a processor and a memory configured to perform the following operations: determine one of the spectral signals obtained for the azimuths Difference signal; if the difference signal is above a predefined threshold, determine and report that the particular structure does not contain a defect; and if the difference signal is equal to or below the predefined threshold or equal to zero , It is determined and reported that the specific structure contains a defect, wherein the defect is lower than a surface of the specific structure. If the system of claim 11, wherein a model is not used and a certain amount of feature values are extracted from this specific structure, a determination is made as to whether the specific structure contains or does not contain a defect. The system of claim 11, wherein the difference signal is an average difference between a plurality of differences between the spectral signals at the azimuth angles through a plurality of wavelengths. The system of claim 11, wherein the difference signal is the highest of the plurality of differences between the spectral signals of the azimuths specified in one of the plurality of wavelength ranges. If the system of claim 13, the processor and the memory are configured to: determine theoretical or measured spectral differences for the azimuths of a non-defective grating structure; and use the theoretical spectral differences Value to normalize the average difference to determine a number of defects. The system of claim 11, wherein measuring the spectrum comprises: generating a differential model by using a two-dimensional beam profile reflectometer. If the system of claim 16, the processor and the memory are configured to determine a difference between the differential model and the best fit between an image and a radially symmetric image. The system of claim 11, wherein the specific structure is selected from a self-directed self-assembled (DSA) structure, an underlying non-directed self-assembled (non-DSA) structure, or a patterned photoresist structure. If the system of item 11 is requested, the processor and the memory are configured to: collect reference data using a CD SEM tool to quantify pattern defects; from the pattern structure of a training set with one of the known pattern defects, according to the Obtain a spectral signal at the same azimuth; determine a first relationship function between the spectral signal measured at different azimuths and a residual error, where the first relationship is based on the azimuth from the training set Obtaining the spectral signals; determining a second relationship function between a residual error and a quantization of a pattern defect based on the reference data; and the spectral signals to be measured from the specific structure according to the azimuth angles Input to the second relationship function to determine a quantification of one of the pattern defects for this particular structure. The system of claim 19, wherein the first relationship function and the second relationship function are based on a data reduction technique applied to the spectral signal and one of the residual errors for the training set and the specific structure.